Priyanka1218/my-stablelm-zephyr-finetune

license: apache-2.0 base_model: stabilityai/stablelm-zephyr-3b tags:

text-generation

fine-tune

qlora

instruct

stablelm

pytorch

My StableLM Zephyr Fine-tune This is a fine-tuned version of the stabilityai/stablelm-zephyr-3b model, trained using the QLoRA method with the Axolotl framework.

Model Description This model was fine-tuned for instruction-following on a small, custom dataset. The training objective was to demonstrate the fine-tuning process and to produce a model that can follow basic instructions and generate text in a clear, concise manner.

Training Details The model was trained for 50 steps using the following key parameters:

Base Model: stabilityai/stablelm-zephyr-3b

Adapter Method: QLoRA

Training Dataset: A small, custom, in-line dataset.

LoRA Rank (lora_r): 32

LoRA Alpha (lora_alpha): 16

Sequence Length: 2048

Optimizer: paged_adamw_32bit

Training Hyperparameters The following hyperparameters were used during training:

learning_rate: 2e-05

train_batch_size: 1

eval_batch_size: 1

seed: 42

gradient_accumulation_steps: 2

total_train_batch_size: 2

optimizer: Use OptimizerNames.PAGED_ADAMW_8BIT with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments

lr_scheduler_type: cosine

lr_scheduler_warmup_steps: 2

training_steps: 25

Evaluation Results The training job completed with the following metrics:

Training Loss: 1.4527

Training Runtime: 940.66 seconds

Training Samples per Second: 0.409

How to Use You can use this model with the transformers library for text generation. Make sure you have the necessary libraries installed and a GPU available.

import torch from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "Priyanka1218/my-stablelm-zephyr-finetune" tokenizer = AutoTokenizer.from_pretrained(model_id) model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float16, device_map="auto")

prompt = "Explain the difference between a list and a tuple in Python." messages = [ {"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": prompt} ] input_ids = tokenizer.apply_chat_template(messages, return_tensors="pt").to("cuda")

outputs = model.generate(input_ids, max_new_tokens=256) print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Dataset The model was fine-tuned on a very small custom dataset in the Alpaca format. For better performance, a larger and more diverse dataset would be required